[
https://issues.apache.org/jira/browse/HADOOP-153?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sharad Agarwal updated HADOOP-153:
----------------------------------
Attachment: 153_1.patch
Here is the initial patch, intended for early feedback. It does the core work
of maintaining the failed ranges and skipping those on further executions. Few
test cases have also been included.
Although the basic functionality is working, the patch is still *incomplete*.
Things remaining are:
- Capture offsets along with record sequence no. Once we have all the offsets
for the skipped records, actual records data can be written anytime later on.
- Write the skipped records information (only sequence nos and offsets) to job
history.
- Support for Pipes and Streaming.
- Introduce thresholds to limit the no of failing tasks execution (failing fast
for really bad user code)
- Execute the skipped ranges on best effort basis. Perhaps this can be done as
a separate Jira later on, if we find value in it after seeing real use cases.
> skip records that throw exceptions
> ----------------------------------
>
> Key: HADOOP-153
> URL: https://issues.apache.org/jira/browse/HADOOP-153
> Project: Hadoop Core
> Issue Type: New Feature
> Components: mapred
> Affects Versions: 0.2.0
> Reporter: Doug Cutting
> Assignee: Sharad Agarwal
> Attachments: 153_1.patch, skipRecords_wip1.patch
>
>
> MapReduce should skip records that throw exceptions.
> If the exception is thrown under RecordReader.next() then RecordReader
> implementations should automatically skip to the start of a subsequent record.
> Exceptions in map and reduce implementations can simply be logged, unless
> they happen under RecordWriter.write(). Cancelling partial output could be
> hard. So such output errors will still result in task failure.
> This behaviour should be optional, but enabled by default. A count of errors
> per task and job should be maintained and displayed in the web ui. Perhaps
> if some percentage of records (>50%?) result in exceptions then the task
> should fail. This would stop jobs early that are misconfigured or have buggy
> code.
> Thoughts?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
